[PATCH 4/4] ARM: keystone: Use dma-ranges for dma_pfn_offset configuration

[Grygorii Strashko]

Hi Arnd,

On 02/24/2014 11:38 PM, Santosh Shilimkar wrote:
> On Monday 24 February 2014 04:11 PM, Arnd Bergmann wrote:
>> On Monday 24 February 2014 15:53:55 Santosh Shilimkar wrote:
>>> From: Grygorii Strashko <grygorii.strashko at ti.com>
>>>   }
>>>   
>>> +static unsigned long get_dma_pfn_offset(struct device *dev)
>>> +{
>>> +	struct device_node *node = of_node_get(dev->of_node);
>>> +	const u32 *ranges = NULL;
>>> +	int len, naddr, nsize, pna;
>>> +	dma_addr_t dma_addr;
>>> +	phys_addr_t cpu_addr, size;
>>> +	unsigned long dma_pfn_offset = 0;
>>> +
>>> +	if (!node)
>>> +		return 0;
>>
>> Hmm, isn't this function the same as of_translate_dma_address()?
>>
>> I think we should have the implementation in common code, not hidden
>> in the keystone platform, to avoid duplication. If of_translate_dma_address
>> doesn't work, what is the problem, and can you fix it there?
>>
> Will have a look at it and get back. At least it makes sense to
> use/update the common function.

This function isn't the same as of_translate_dma_address(), more over
it calls of_translate_dma_address() to get CPU address.

To calculate DMA bus offset we need to know the first DMA address explicitly,
so then we can translate it to the corresponding CPU address.

get_dma_pfn_offset() function does:
- traverse DT and look for the first "dma-ranges" prop;
- reads the first DMA address of the range;
- translates it in CPU address using of_translate_dma_address();
- calculates DMA bus offset as PFN_DOWN(cpu_addr - dma_addr);

Also, I've investigated all places in Kernel were "dma-ranges" is used:
- cell_iommu_get_fixed_address() - arch/powerpc/platforms/cell/iommu.c
- ppc4xx_parse_dma_ranges() - arch/powerpc/sysdev/ppc4xx_pci.c

As result, I see no reasons to modify of_translate_dma_address()
(which actually is just wrapper for common __of_translate_address()).

Of course, possibly I've not fully got your point. :)

> 
>>>   static int keystone_platform_notifier(struct notifier_block *nb,
>>>   				      unsigned long event, void *dev)
>>>   {
>>> +	struct device *_dev = dev;
>>> +
>>>   	if (event != BUS_NOTIFY_ADD_DEVICE)
>>>   		return NOTIFY_DONE;
>>
>> Style: it would be nicer to name the local variable 'dev' and the
>> argument something else like 'p' or 'data'.
>>
> Agree. Will update that.
>   
>> I also wonder if this shouldn't be in ARM architecture wide code
>> rather than platform code. Unfortunately it can't be in drivers/base
>> since the offset is stored in an ARM specific location.
>>
> Notifier callback is in mach code mainly because platform's might have to
> do custom things here. Like setting up the coherent masters,
> populating the dma_pfn_offset or populating custom dma_ops etc.
> As such the in these callbacks is not much and I see only
> couple of machines using it.
> 

Thanks for your comments.

Regards,
-grygorii